Application Of Decision Level Information Fusion For The Fault Diagnosis Of Suction Fan

The suction fan is important equipment in the factory, based on the factory's real condition and management request, in order to ensure it run safely and successfully, it is necessary to research the diagnostic system of the suction fan fault. Recently, the technology of fault diagnosis is developing very fast and there are many methods for it, but most of them are the simplex methods. Based on analyzing the currently domestic and foreign study, a fusion diagnostic system based on the BP neural network and the synthetically relational analysis is presented. The system is tested by the sample of the data collected from different placements. The main work is as follows:After analyzing the suction fan fault and the historical data, according to the experience of expert, eight normative faults with sample data is concluded.This paper also analyzes the fault diagnosis based on the relational analysis. In the relational analysis, the traditionally relational degree only considers the similar characteristic not the distance between the fault modes. For this reason, the similar coefficient in cluster analysis is introduced to form the synthetically degree with traditionally relational degree. The experiment indicates this method is available for the fault diagnosis of suction fan.The basic probability assignment function of D-S evidence theory is presented by the output of the neural network and the synthetically degree. The partial fusion diagnosis system based on the BP neural network and D-S evidence is established, so does the partial fusion diagnosis system based on the synthetically degree and D-S evidence. The decision fusion is taken in these two methods. The result indicated that this method can reduce the uncertainty of decision and greatly increase the precision of diagnosis.